• Home
• Program
• Contact

10:45   Numerical Methods and Data Analysis 1 10:45 15 mins #372 Temporal Large-Eddy Simulation with exact deconvolution Daniel Oberle, David Pruett, Patrick Jenny Abstract: We propose an approach for Temporal Large-Eddy Simulation (TLES) with Temporal Exact Deconvolution (TEDM). In contrast to previous approaches like the Temporal Approximate Deconvolution Model (TADM) by Pruett et al. [3], the nonfiltered field is recovered using the differential form of the exponential filter rather than a truncated series expansion of the inverse filter operator. The closure is based on an analytic evolution equation of the temporal stress tensor. Furthermore, a different regularization term based on selective frequency damping is used. The model is used to simulate incompressible test cases such as a turbulent channel with Reτ=180 and homogeneous isotropic turbulence with Reλ=190 using the spectral element code NEK5000 [1]. As seen in Figs. 1 and 2 the results demonstrate that the model significantly improves the results compared to the no-model solution while saving computational cost in comparison to direct numerical simulation. The main advantage of TEDM is that it provides a natural framework to couple LES with RANS simulations. Figure 1: Turbulent channel with Re τ = 180: Mean velocity profile in wall units, DNS by Kim et al. 1998 [2] with N y =129, coarse DNS with N y =49, TLES with N y =49, where N y is the number of grid points in the y-direction. Figure 2: Energy spectrum of homogenous isotropic turbulence with Re λ = 190: , DNS with N =176, coarse DNS with N =36, TLES with N =36, where N is the number of grid points in each direction. References [1] P. Fischer. Features of NEK5000. [2] J. Kim, P. Moin, and R. Moser. Turbulence statistics in fully developed channel. Journal of Fluid Mechanics, 177:133–166, 1987. [3] C. D. Pruett, B. C. Thomas, C. E. Grosch, and T. B. Gatski. A temporal approximate deconvolution model for large-eddy simulation. Physics of Fluids, 18(2):1–5, 2006. 11:00 15 mins #565 On a proper tensor-diffusivity model for large-eddy simulations of Rayleigh-Bénard convection F.Xavier Trias, Firas Dabbagh, Daniel Santos, Andrey Gorobets, Assensi Oliva Abstract: In this work, we plan to shed light on the following research question: can we find a nonlinear tensorial subgrid-scale (SGS) heat flux model with good physical and numerical properties, such that we can obtain satisfactory predictions for buoyancy driven turbulent flows? This is motivated by our recent findings showing that the classical (linear) eddy-diffusivity assumption fails to provide a reasonable approximation for the SGS heat flux. We have also concluded that nonlinear (or tensorial) models can give good approximations of the actual SGS heat flux. The nonlinear Leonard model is an example thereof. However, this model is unstable and therefore it cannot be used as standalone SGS heat flux model. Apart from being numerically stable we also want to have the proper cubic near-wall behavior. Corrections in this regard will be presented together with a priori/posteriori studies of nonlinear SGS heat flux models for RBC. Results from LES simulations will be compared with the DNS results obtained in the recent PRACE project 'Exploring new frontiers in Rayleigh-Bénard convection'. 11:15 15 mins #36 Controlled eddy simulation of complex wall bounded flows at large Reynolds numbers Yan Jin Abstract: Turbulent flows are often observed phenomena in everyday surroundings and prevalent processes in industry. Simulation of turbulent flows, particularly at high Reynolds numbers or with a complex geometry of bounded walls, is a challenging task. Controlled eddy simulation (CES) as a new method for simulating turbulent flows has been proposed in the study. An artificial force is introduced into the momentum equation in order to filter part of the turbulent motions. A weight coefficient is used to determine the number of eddies to be filtered. The model equations of CES approach to the DNS equations as the weight coefficient is reduced to zero. Therefore, a CES solution can be seen as an asymptotic approximation of the DNS solution. A modified mixing length (ML+) turbulence model has been developed for CES. The CES method was used to calculate two types of complex wall bounded flows at high Reynolds numbers. The first test case is about turbulent flows in a channel with d-type rough walls. The Reynolds numbers based on the friction velocity Reτ up to 5600 have been accounted for in the simulation. The CES results for the friction coefficient, mean velocity, and Reynolds stresses were compared with our DNS and RANS solutions in [1]. The roughness function was determined from the CES solution. The second test case is about turbulent flows in a compressor cascade. The Reynolds number based on the airfoil chord is 3.82×105. Large eddy simulation (LES) of the same flow requires about 4×108 grid points, see [2]. Only 1.6×107 grid points were used for CES. The CES results for the loss and pressure coefficients were validated by the experimental data. The mean features of cascade flows, e.g., laminar-turbulence transition, boundary layer separation, secondary flows, can be clearly identified from the CES solution, see Fig. 1. The numerical results show that CES is an efficient and accurate method for simulating turbulent flows. It is particularly suitable for the flows at high Reynolds numbers with complex geometries of bounded walls. Profound flow details and statistical results can be obtained from the CES solution, while its computational cost is much lower than that of LES. Due to the flexibility for developing CES models, there is still much room for further improvement of the model accuracy and reduction of the computational costs. 11:30 15 mins #607 Effects of spatial filtering on scale-to-scale energy flux Daniel Feldmann, Jan Chen, Marc Avila Abstract: Here, we examine the scale-to-scale flux of turbulent kinetic energy between different length scales in a turbulent pipe flow based on an adapted approach by Eyink and co-workers, which relies on spatial filtering. Three different spatial filter approaches — a Fourier space cut-off filter, a physical space Box filter, and a smooth Gauss filter — were implemented in a three-dimensional cylindrical coordinate frame work using the interpreted programming language Python. To incorporate the inhomogeneous wall-normal direction, various combinations of different boundary conditions (BC) as well as one-, two-, and three-dimensional filters were realised and systematically compared to each other. These filters were applied to our data base of highly resolved velocity fields from a direct numerical simulation of a low-Reynolds-number turbulent pipe flow using our pseudo-spectral Navier-Stokes solver. Based on the filtered velocity fields, we computed for the first time the inter-scale energy flux in a wall-bounded turbulent flow based on three-dimensional spatial filtering in a cylindrical coordinate frame work. At the conference, we will discuss the influence of the type of filtering on the computed flux of energy and their physical interpretation. 11:45 15 mins #92 Development and Investigation of Thermal Subgrid-Scale Models for Large Eddy Lattice Boltzmann Methods Maximilian Gaedtke, Daniel Rau, Hermann Nirschl, Mathias J. Krause Abstract: The increasing speed of modern high-performance computers allows the numerical simulation of more and more complex fluid flow problems. Turbulent flows caused by thermal buoyancy forces pose a particular challenge for the simulation, since the often strongly transient fluctuations are difficult to model using Reynolds Averaged Navier Stokes (RANS) approaches. However, such models are essential to predict weather phenomena more precisely, to develop more efficient insulating materials, latent heat storage or process engineering equipment. Previous work has shown that Lattice Boltzmann Methods (LBM) are particularly suitable for Large Eddy Simulations (LES). They make very efficient use of parallel computers with several thousand CPU cores or GPUs and allow simulations on large computing grids with hundreds of millions of cells. Although some subgrid-scale models for the description of thermal flows under the influence of buoyancy forces are proposed in the literature, they are mainly used within the framework of the Finite Volume Method. Neither Lattice Boltzmann discretizations of the models according to Eidson [1], Peng and Davidson [2] nor the thermal mixed scale model according to Sergent et al. [3] are currently available. The aim of this work is therefore to derive, implement, test and evaluate LB-based approaches for these models, while maintaining the locality and the associated parallelizability of the LB algorithm. For validation, mean values of temperatures and velocities as well as common turbulent statistics are compared with high accuracy measurement results and simulation data of other authors obtained for the benchmark case "natural convection in a air-filled differently heated cavity", whereby the strengths and weaknesses of the respective turbulence model as well as its implementation by means of LBM are presented in detail. 12:00 15 mins #489 Spectral Simulations of Quantum Turbulence using the Gross-Pitaevskii Equation Michikazu Kobayashi, Ionut Danaila, Corentin Lothode, Francky Luddens, Philippe Parnaudeau, Luminita Danaila Abstract: We present spectral numerical simulations of the Gross-Pitaevskii (GP). We simulate the dynamics of Quantum Turbulence (QT) superflows for two distinct cases: (i) periodic box without trapping potential and rotation and (ii) periodic box with confining (harmonic) potential and rotation around a major axis. The former case corresponds to the classical setting GP-QT simulations of superflows (e.g. superfluid 4He), while the latter corresponds to more recent settings of QT in Bose-Einstein condensates (BEC). After validating the code using initial conditions based on Taylor-Green vortices or random-phase fields, we present a new type of initial condition based on randomly generated quantum vortex rings. We will discuss in detail the characteristics of the QT generated with these three different initial conditions 12:15 15 mins #454 A GENERAL FORMALISM FOR SCALES INTERACTION AND THEIR MODELLING IN LES Antonella Abbà, Andrea Cimarelli, Andrea Crivellini, Massimo Germano Abstract: Starting from an alternative decomposition of the subgrid stresses, we present a tensorial turbulent viscosity for a reduced description of small-scale turbulence. This formulation highlights the relevance of taking into account how the velocity field is distributed in the three spatial directions within the computational volume and, hence, it could be understood as the basis for the development of turbulence models. Here we propose a model based on inertial properties of the grid elements. The basic idea is that the structure of the mesh reflects the anistropy and inhomogeneity of the flow and could be used to predict the main features of small scale turbulence. This aspect is particularly relevant in real-world problems where the complexity of the unstructured meshes usually adopted strongly reflects the complexity of the flow dynamics. When cartesian grids are used, the present approach is recognized to recover the subgrid viscosity of the well-known gradient model. On the other hand, we may expect that the present formulation should improve the LES approach when complex flow system characterized by unstructured grid are considered.